Dynamic Checkpoint Initiation in Serverless MEC

Abstract

Mobile applications may want to offload heavy processing jobs to more powerful nodes. Mobile Edge Computing (MEC) can provide low latency and fast processing for the offloaded jobs. Due to the limited capacity of MEC, an efficient checkpointing mechanism can provide fair resource sharing among offloaded jobs by interrupting them as necessary. It is important for an MEC controller to decide when to start checkpointing based on the resource availability and the number of queued requests. In this paper, we propose a resource management framework that decides when to start checkpointing to utilize the MEC compute resources efficiently. We also show how we can integrate the proposed framework in practical implementation of an MEC architecture.